| CPC G06V 40/28 (2022.01) [G06F 3/017 (2013.01); G06T 7/251 (2017.01)] | 20 Claims |

|
1. A three-dimensional gesture detection device, comprising:
a node detection unit, used to receive a continuous hand image containing a plurality of hand frames, wherein the node detection unit obtains a plurality of nodes according to each of the hand frames;
a gesture recognition model, used to obtain a plurality of confidence levels of a plurality of the gesture categories each of which is tap, pick, open the hand, grip, or raise the hand according to the nodes; and
a gesture trajectory detection unit, comprising:
a weight analyzer, used to obtain a plurality of weights of the gesture categories each of which is tap, pick, open the hand, grip, or raise the hand through a user interface;
a gesture analyzer, used to performs a weighting calculation on the confidence levels of the gesture categories according to the weights to analyze a gesture on each of the hand frames;
a key point classifier, used to classify a plurality of key points from part of the nodes according to the gesture; and
a trajectory analyzer, used to obtain an inertial trajectory of the gesture according to changes in the key points on the hand frames.
|